Developing Scale-Aware Algorithms for the Ensemble Filtering of Geophysical Flows

Abstract

Geophysical flows typically contain multiple scales of motion. The larger scales tend to have slower and more linear error growth than the smaller scales. With smaller scales being resolved in numerical models as computational resources increase, their nonlinear behavior poses a challenge to linear ensemble filters. In this seminar, I will discuss several issues that occur when applying an ensemble filter to a multiscale dynamical system and explores potential remedies. The sensitivity of the best Gaspari-Cohn (GC) localization radius to changes in model resolution, ensemble size and observing networks is systematically evaluated with assimilation experiments using a two-layer quasi-geostrophic (QG) model. Results demonstrate the scale dependency of the best localization radius for the QG flow. The best localization radius is sensitive to the underlying physical correlation length scale but is much less sensitive to model resolution as long as the key physical process is well represented by a truncated model. The GC localization is problematic when dealing with nonlocal cross-variable correlations, which motivates the use of localization functions that respect the dynamic relations between variables. The impact of varying observing network density on the best choice of localization is also discussed.

YUE (MICHAEL) YING

National Center for Atmospheric Research